stripping off "xf86-*-" from drivers

[Dave Airlie]

On Jan 21, 2008 3:10 PM, Luc Verhaegen <libv at skynet.be> wrote:
> On Mon, Jan 21, 2008 at 02:33:16PM +1000, Dave Airlie wrote:
> >
> > Nobody said we should make it impossible to build drivers out of tree
> > pr throw away the SDK, just that the default state for most drivers
> > should be in the tree. the kernel doesn't make it impossible to build
> > drivers out of tree at all, its just not a task you want to do for
> > long, however I do it for the DRM mostly and have no major problems..
> >
> > Dave.
>
> So... You're certain that something like libpciaccess wouldn't have
> happened as such, if this part of the X was still a monolith? And would
> it have happened as it did if there was a working tinderbox?

There would've been a lot less chance, I don't think you can ever be
"certain" without doing it. I don't think a working tinderbox would've
helped as the input api breakage would've set the tinderbox red long
before then.. now if we'd fixed all the input drivers.. recurse to
previous API change etc... we did have a working tinderbox once.

> I'm sure that we can all see what will happen to the SDK, with what you
> propose.
>
> But guess what, you can forget about giving people easy access to new
> hardware support on a distribution that is currently shipping.

Luc some of us do this for a lot more customers and with a lot more
success every day.. maybe it'll get harder for us in the future, but
distro vendors shouldn't be setting all the direction here...

> When people have been using vesa for months, for lack of a proper
> driver, or when people have just bought a new graphics card, their only
> option suddenly becomes to upgrade the whole X server stack.

Come on, you work for a vendor.. does this happen with your kernels?

Dave.